Most "AI for data science" tools stop at a chat box that writes pandas snippets you still have to wire together. This project takes the opposite bet: it ships a roster of narrow, single-job agents — one for cleaning, one for feature engineering, one for SQL, one for H2O modeling — so the LLM's job becomes orchestration rather than reinventing each step from a blank prompt.
What Sets It Apart
- Each agent emits real, inspectable Python or SQL, not just a chat answer — you can lift the generated code into a notebook or pipeline and rerun it without the model in the loop.
- Coverage spans the whole workflow: data loading, wrangling, cleaning, visualization, EDA, feature engineering, and SQL databases, plus H2O AutoML and MLflow tracking agents.
- Multi-agent analysts (Pandas Data Analyst, SQL Data Analyst) and a Supervisor Agent compose several specialized agents into a single task, instead of you calling each one by hand.
Great Fit / Look Elsewhere
Great fit if you already live in the Python data stack (pandas, scikit-learn, H2O) and want LLM agents that hand back reproducible code for routine, repetitive steps. Look elsewhere if you need a no-code BI dashboard, a general-purpose coding assistant, or production guarantees — this is an evolving, code-first toolkit where you stay on the hook for reviewing whatever each agent generates before trusting it.
